MintWrite AI

AI-powered content marketing platform for Web3 projects. Generates Twitter threads, Discord announcements, tokenomics explainers, blog posts, AMA scripts, and whitepaper summaries in under 60 seconds.

Quick start

npm install
cp .env.example .env.local
# Fill in your keys (see Environment Variables below)
npm run dev

Open http://localhost:3000

Environment Variables

Variable	Required	Description
ANTHROPIC_API_KEY	✅	Claude API key
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY	✅	Clerk publishable key
CLERK_SECRET_KEY	✅	Clerk secret key
POSTGRES_URL	✅	Vercel Postgres URL
LEMON_SQUEEZY_API_KEY	✅	Lemon Squeezy API key
LEMON_SQUEEZY_STORE_ID	✅	Your LS store ID
LEMON_SQUEEZY_WEBHOOK_SECRET	✅	Webhook signing secret
LEMON_SQUEEZY_PRO_VARIANT_ID	✅	Pro plan ($49/mo) variant ID
LEMON_SQUEEZY_AGENCY_VARIANT_ID	✅	Agency plan ($149/mo) variant ID

Database setup

Run the SQL schema in your Vercel Postgres dashboard:

psql $POSTGRES_URL -f lib/db/schema.sql

Project structure

app/
  (auth)/          # Clerk sign-in / sign-up pages
  (app)/           # Protected app routes
    dashboard/     # Content Hub — 7 content type cards
    generate/[type]/ # Generation form + streaming output
    history/       # Past generations
  api/
    generate/      # Streaming AI endpoint (maxDuration: 120)
    projects/      # Project CRUD
    generations/   # Save & list generation history
    webhooks/lemon/ # Lemon Squeezy payment webhooks
  page.tsx         # Landing page
  pricing/         # Pricing table

lib/
  ai/
    content-types.ts  # Registry of all 7 content types (add new types here)
    prompts/          # System prompt + 7 user prompt builders
    schema.ts         # Zod validation schemas
  db/
    client.ts         # Vercel Postgres queries
    schema.sql        # Table definitions
  payments/lemon.ts   # Checkout URLs + webhook verification
  utils/tier.ts       # Tier capability checks

components/
  forms/
    GenerationForm.tsx   # Dynamic form driven by content type registry
    ProjectSetupForm.tsx # One-time project setup
  output/
    ContentOutput.tsx    # Streaming output display
    CopyButton.tsx       # Copy with confirmation feedback

Adding a new content type

1. Create lib/ai/prompts/your-type.ts with buildYourTypePrompt(project, context)
2. Add the type definition to CONTENT_TYPES in lib/ai/content-types.ts
3. Add the prompt builder to PROMPT_BUILDERS in the same file

Nothing else changes. The UI, API, and form are data-driven from the registry.

Architecture decisions

Streaming API — uses Vercel AI SDK streamText with maxDuration: 120 to avoid Vercel's 10s serverless timeout. The client reads the stream and renders output in real time.

Content type registry — all 7 types are defined as data objects, not as code branches. Adding a new type requires zero changes to UI or API.

Tier checks — lib/utils/tier.ts is the single place where tier capabilities are defined. Server-side checks in API routes; client-side checks for UI gating only.

CleanCode / SOLID — single-responsibility modules, no magic strings, explicit return types, no any, Zod validation on all external data.

Related products

• TokenForge AI — Tokenomics design tool (same customer, cross-sell opportunity)